Your people are already using AI. The question is whether it happens inside your system.
Government and sector guidance encourages careful, deliberate adoption - not avoidance. The risk that keeps coming up in our conversations with customers is 'shadow IT': staff and applicants pasting sensitive material into public tools because there is no sanctioned alternative. We built Omni AI so there is one.
Guidance is clear
Governments and funding bodies around the world have published frameworks for adopting AI responsibly, and most funders now set expectations for how AI may be used in applications and assessment. We have mapped Omni AI against them.
Applications keep growing
Rounds attract more applications every year, and reviewers have the same number of hours. Omni AI lifts the quality of what arrives and gives reviewers a second perspective on what they receive.
You keep the decision
Omni AI informs the decision-making process. Your staff make the call, every time. That is a design principle, not a disclaimer - and the event log proves it.
Our goal has been to provide useful functionality which streamlines the work of users by enhancing capacity and capability with advanced help and valuable insights - without compromising the privacy and security of your data.
The OmniStar product teamFour features, two jobs
Omni AI works alongside OmniStar's core functionality across the whole lifecycle. Its features group into help and insights - and each one is deliberately scoped to a job we know research offices and grant teams do every day.
Interactive help
A context-aware assistant that knows where you are in OmniStar. Ask a question in plain language and get a tailored answer drawn from our wiki - including worked examples, not just a link to the page that explains the concept.
Ask for what you need next
Our roadmap includes follow-up questions in reviews, customised prompts, insights informed by the context of the round, and bias prevention. We are building these with customers, not for them.
Application insights
Applicants can critique their own draft before they submit. Omni AI highlights gaps, inconsistencies and areas to strengthen - considering both the application and any custom instructions you set for the round. Fewer revision cycles, higher quality proposals.
Review insights
Our roadmap includes follow-up questions in reviews, customised prompts, insights informed by the context of the round, and bias prevention. We are building these with customers, not for them.
Document insights
Ask whether a question is answered anywhere across an application, project, milestone submission, variation request or agreement. "Does this project include an ethics approval pathway?" - Omni AI finds where it is mentioned, or tells you plainly that it is not.
We recognise the responsibility that comes with building this
Five principles guide every part of our AI implementation. They are the standard we hold ourselves to, and they are what we will be judged against.
We enhance human capability, not replace it
Omni AI supplements your staff and users. It provides insight and streamlines workflows while your people remain the final decision makers.
We embrace innovation and co-design
We work with customers to solve their challenges by evolving our AI capabilities to suit their needs.
We prioritise security and privacy
We use secure, locally hosted AI models that protect your data. We do not trade privacy for convenience.
We ensure fairness and ethical behaviour
We select AI models with strong anti-bias testing, we welcome feedback, and we constantly enhance our features to better serve you.
We maintain transparency and accountability
We clearly label AI features, implement event logging, track the history of generated insights, and help you ensure AI-influenced outcomes are reversible.
Safe use of AI is people, process and technology
Developed in response to leading government AI assurance guidance, our shared responsibility model sets out exactly what we take care of and what you should consider doing. No ambiguity about where the line sits.
| Training | Privacy & security | Fairness | Accountability | Transparency | Operation & feedback | |
|---|---|---|---|---|---|---|
| OmniStar | Provide AI documentation | Secure operation of the model Model training disabled | Select fair, unbiased AI models | Audit log of AI usage Maintain history of insights | Get consent to switch Omni on Notify users when using AI features | Collect and action feedback |
| Customer | Train users | Consider a privacy impact assessment | Apply AI consistently | Implement an AI use policy Decision-making responsibility | Notify users that AI is used in the round or organisation | Manage residual risks Test features and provide feedback |
Policies have little use if it cannot be proven that they were followed. Our event log shows that Review Insights was used - and validates that it was used fairly, because it was used on every application.
From OmniStar Insights: using AI responsiblyProtected by our technical implementation - not just a policy statement
The most common concern we hear is that sensitive data, whether provided directly or derived from the questions you ask, gets harvested and fed back into the model. Here is exactly why that does not happen with Omni AI.
Inference, not training
Training is the years of study before professional starts work. Inference is consulting that professional. Your data is processed solely to generate a response - it is never incorporated into training datasets and never used to enhance the model's capabilities.
Your data stays in your country
Omni AI runs on Amazon Bedrock, with all operations conducted entirely within your AWS hosting country. Your data never leaves your country's shores, giving you both data sovereignty and compliance with local regulations.
Isolated model deployment
Models are deployed in dedicated accounts controlled by our infrastructure provider's service team, not by the model providers. That is a clear separation between the operator of the model and its trainer.
Enterprise-grade encryption
TLS 1.2 in transit and AES-256 at rest, the same standard applied across the rest of OmniStar. Only authorised OmniStar personnel can access the system.
Transparent data handling
Amazon Bedrock does not store or log your prompts and responses. OmniStar keeps secure records of these interactions for accountability, so you can see how AI influenced a decision if a decision is ever appealed.
Our model partners
We use Anthropic's Claude Sonnet 4.5 and Cohere's Embed English, selected for their performance. Our implementation ensures these providers cannot access your data, cannot use it for training, and have no visibility into your interactions.
Trusted by industry leading organisations
Omni AI is off until you turn it on
Omni AI operates on a controlled access model. Every feature is disabled by default, activation is explicit, and consent is required at both the organisation and the individual user level. You decide what is switched on, for whom, and when.
Consultation
Before anything is switched on, our team runs a session to assess your requirements, address your specific concerns, and agree how Omni AI should be configured for your rounds.
Organisational activation
AI features are disabled by default for every organisation and require explicit activation. Nothing turns on by surprise, and nothing turns on without your consent.
Individual consent
Once your organisation has enabled Omni AI, each user must give their own consent before accessing any AI capability. Two tiers, both under your control.
Feedback and support
Dedicated staff support you throughout, and we escalate any issue you raise - ultimately switching Omni AI features off if that is what is needed.
Frequently asked questions
Omni AI is a collection of machine learning-assisted features within OmniStar, the most powerful end-to-end solution for grants, ethics and research management. It streamlines the work of applicants, reviewers and administrators with advanced help and valuable insights, without compromising the privacy and security of your data.
No. Omni AI informs the decision-making process; your staff always make the final decision. This ensures your people remain accountable to applicants and to the communities that benefit from funding, and it protects against model bias and hallucination. We recommend documenting decision-making responsibility in your AI use policy so your communities understand your intent.
No. Your data is processed for inference only, and is never incorporated into training datasets or used to enhance the model's capabilities. This is enforced by our technical implementation, not by a policy statement alone: models are deployed in isolated accounts, and our model providers cannot access your data, cannot use it for training, and have no visibility into your interactions with Omni AI.
Entirely within your AWS hosting country, through Amazon Bedrock. Your data never leaves your country's shores, supporting both data sovereignty and compliance with local regulations.
Our event log records the use of AI features and can be used as evidence. For example, it can show that Review Insights was applied to every application in a round rather than to some but not others - which is how you demonstrate that availability bias was avoided.
We also retain a history of the insights Omni AI generated, so if an applicant appeals a decision you can see the extent to which an insight informed your staff's judgement.
Anthropic's Claude Sonnet 4.5 and Cohere's Embed English. We selected them for their performance and because they align with our principles, including verifiable fairness measures such as scoring well on bias testing.
We produce detailed documentation that maps Omni-specific information to the assurance frameworks our customers work under, covering why AI is suited to the task, community benefit, fairness, privacy and security, transparency, accountability and risk management. It is available on request and is designed to help you complete your own assessment.
We have also completed a generic privacy impact assessment, which you are welcome to use as the basis for your own.
Speak to your Account Manager, or book a demo below. We start with a consultation to understand your requirements and concerns, then enable the features you have chosen at organisation level. Your users each give their own consent before they can use them, and we support you throughout - including switching features off again if that is what you need.
Ready to book a demo?
Book with us today and unlock the power of OmniStar.